AFrameworkandInfrastructureforSemanticWebServices
EnricoMotta1,JohnDomingue1,LilianaCabral1,andMauroGaspari2
1KnowledgeMediaInstitute,TheOpenUniversity,MiltonKeynes,UK{
E.Motta,
J.B.Domingue,
L.S.Cabral}@open.ac.uk 2DipartimentodiScienzedell’Informazione,UniversityofBologna,Italygaspari@cs.unibo.it Abstract.InthispaperwedescribeIRS–II(ReasoningService)aframeworkandimplementedinfrastructure,whosemaingoalistosupportthepublication,location,positionandexecutionofheterogeneouswebservices,augmentedwithsemanticdescriptionsoftheirfunctionalities.IRS–IIhasthreemainclassesoffeatureswhichdistinguishitfromotherworkonsemanticwebservices.Firstly,itsupportsone-clickpublishingofstandalonesoftware:IRS–IIautomaticallycreatestheappropriatewrappers,givenpointerstothestandalonecode.Secondly,itexplicitlydistinguishesbetweentasks(whattodo)andmethods(howtoachievetasks)andasaresultsupportscapabilitydrivenserviceinvocation;flexiblemappingsbetweenservicesandproblemspecifications;anddynamic,knowledge-basedserviceselection.Finally,IRS–IIservicesarewebpatible–standardwebservicescanbetriviallypublishedthroughtheIRS–IIandanyIRS–IIserviceautomaticallyappearsasastandardwebservicetootherwebserviceinfrastructures.InthepaperweillustratethemainfunctionalitiesofIRS–IIthroughascenarioinvolvingadistributedapplicationinthehealthcaredomain. 1Introduction Webservicespromisetoturnthewebofstaticdocumentsintoavastlibraryofinteroperableputerprogramsandassuchhaveattractedconsiderableinterest,bothfromindustryandacademia.Forexample,IDC[8]predictsthattheWebServicesmarket,valuedat$416millionin2002,willbeworth$2.9billionby2006. Existingwebservicetechnologiesarebasedonamanualapproachtotheircreation,maintenanceandmanagement.Atthecentreoftheconceptualarchitectureisaregistrywhichstoresdescriptionsofpublishedwebservices.Clientsquerytheregistrytoobtainrelevantdetailsandtheninteractdirectlywiththedeployedservice.Thedescriptions,representedinXMLbaseddescriptionlanguages,suchasWSDL[17]andUDDI[16],mostlyfocusonthespecificationoftheinputandoutputdatatypesandtheessdetails.Thesespecificationsareobviouslynotpowerfulenoughtosupportautomaticdiscovery,mediationpositionofwebservices.Asoftwareagentcannotfindoutwhatawebserviceactuallydoes,byreasoningaboutaWSDLspecification.Analogouslythesameagentcannotlocatetheappropriate
D.Fenseletal.(Eds.):ISWC2003,LNCS2870,pp.306–318,2003.©Springer-VerlagBerlinHeidelberg2003 IRS–II:AFrameworkandInfrastructureforSemanticWebServices307 serviceinaUDDIregistry,giventhespecificationofatargetfunctionality.Asaresult,existingwebserviceinfrastructuresbyandlargesupportamanualapproachtowebservicemanagement:onlymanualdiscoveryissupportedandonly‘static’,manuallyconfiguredwebapplicationsarepossible. Theaboveissuesarebeingaddressedbyongoingworkintheareaofsemanticwebservices[3,5,14].Thebasicideahereisthatbyaugmentingwebserviceswithrichformaldescriptionsofpetencemanyaspectsoftheirmanagementwilleautomatic.Specifically,webservicelocation,positionandmediationcanedynamic,withsoftwareagentsabletoreasonaboutthefunctionalitiesprovidedbydifferentwebservices,abletolocatethebestonesforsolvingaparticularproblemandabletoposetherelevantwebservicestobuildapplicationsdynamically.Researchintheareaisrelativelynewandalthoughanumberofapproacheshavebeenproposed,suchasDAML-S[3]andWSMF[5],prehensivetoolinfrastructuresexist,whichsupportthespecificationanduseofsemanticwebservices. InthispaperwedescribeIRS–II(ReasoningService)aframeworkandimplementedinfrastructurewhichsupportsthepublication,location,positionandexecutionofheterogeneouswebservices,augmentedwithsemanticdescriptionsoftheirfunctionalities.IRS–IIhasthreemainclassesoffeatureswhichdistinguishitfromotherworkonsemanticwebservices. Firstly,itsupportsone-clickpublishingof‘standard’programmingcode.Inotherwords,itautomaticallytransformsprogrammingcode(currentlywesupportJavaandLispenvironments)intoawebservice,byautomaticallycreatingtheappropriatewrappers.Hence,itisveryeasytomakeexistingstandalonesoftwareavailableon,aswebservices. Secondly,theIRS–IIbuildsonknowledgemodelingresearchonponentsforknowledge-basedsystems[2,6,9,10],andasaresult,itsarchitectureexplicitlyseparatestaskspecifications(theproblemswhichneedtobesolved),fromthemethodspecifications(waystosolveproblems),fromthedomainmodels(wheretheseproblems,whichcanbegenericinnature,needtobesolved).Asaconsequence,IRS–IIisabletosupportcapability-drivenserviceinvocation(findmeaservicethatcansolveproblemX).Moreover,thecleandistinctionbetweentasksandmethodsenablesthespecificationofflexiblemappingsbetweenservicesandproblemspecifications,thusallowingan:mmappingbetweenproblemsandmethodsandadynamic,knowledge-basedserviceselection. Finally,IRS–IIservicesarewebpatible–standardwebservicescanbetriviallypublishedthroughtheIRS–IIandanyIRS–IIserviceautomaticallyappearsasastandardwebservicetootherwebserviceinfrastructures. Thepaperanizedasfollows:inthefollowingsectionweoutlineouroverallapproach.WethendescribetheIRS–IIframeworkindetailandillustrateitsponentsthroughascenarioinvolvingadistributedhealthcareapplication.Thefinalsectionofthepapercontainsourconclusions. 2TheIRS–IIApproach WorkontheIRS–IIbeganinthecontextoftheIBROWproject[1],whoseoverallgoalwastosupporton-the-flyapplicationdevelopmentthroughtheautomatic 308E.Mottaetal. configurationofreusableponents,availablefromdistributedlibrariesonthe.TheselibrariesarestructuredordingtotheUPMLframework[6],whichisshowninfigure1.TheUPMLframeworkdistinguishesbetweenthefollowingclassesponents:•Domainmodels.Thesedescribethedomainofanapplication(e.g.vehicles,a medicaldisease).•Taskmodels.Theseprovideagenericdescriptionofthetasktobesolved, specifyingtheinputandoutputtypes,thegoaltobeachievedandapplicablepreconditions.Taskscanbehigh-levelgenericdescriptionsplexclassesofapplications,suchasClassificationorScheduling,aswellasmore‘mundane’problemspecifications,suchasExchangeRateConversion.•ProblemSolvingMethods(PSMs).Theseprovideabstract,implementationindependentdescriptionsofreasoningprocesseswhichcanbeappliedtosolvetasksinspecificdomains.Asinthecaseoftaskmodels,thesecanbehigh-level,genericPSMssuchasHeuristicClassification[2]andPropose&Revise[9],ortheycanbespecializedmethodsapplicabletofine-grainedtasks,suchasExchangeRateConversion.•Bridges.Thesespecifymappingsbetweenthedifferentponentswithinanapplication.Forexample,therefinementprocessinheuristicclassificationmaybemappedontoataxonomichierarchyofattributeswithinsomedomain,inordertoconstructaspecificapplication.Eachclassponentisspecifiedbymeansofanappropriateontology[7].Themainadvantageofthisframeworkfromanepistemologicalpointofviewisthatitclearlyseparatestheponentsofknowledge-basedapplications,thusprovidingatheoreticalbasisforanalyzingknowledge-basedreasonersandanengineeringbasisforstructuringlibrariesofponents,performingknowledgeacquisition,andcarryingoutapplicationdevelopmentbyreuse[10]. Fig.1.TheUPMLframework TheapplicationoftheUPMLframeworktosemanticwebservicesalsoprovidesanumberofadvantagesandinourviewourparesfavorablywithapproachessuchasDAML-S,whereservicesarearrangedinhierarchiesandnoexplicitnotionoftaskisprovided–tasksaredefinedasservice-seekingagents.In IRS–II:AFrameworkandInfrastructureforSemanticWebServices309 DAML-Stasksarealwaysapplicationspecific,noprovisionfortaskregistriesisenvisaged.Incontrastinourapproach,tasksprovidethebasicmechanismforaggregatingservicesanditispossibletospecifyservicetypes(i.e.,tasks),independentlyofspecificserviceproviders.InprinciplethisisalsopossibleinDAML-
S.Hereataskwouldbedefinedasaserviceclass,sayS,anditsprofilewillgivethetaskdefinition.However,thissolutionimpliesthatallinstancesofSwillinheritthetaskprofile.Thisapproachisnotveryflexible,giventhatitmakesitimpossibletodistinguish(andtoreasonabout)thedifferencesbetweentheprofileofatask(serviceclass)andtheprofileofamethod(specificserviceprovider)–attributesareinheriteddownis-ahierarchies.Inparticular,insomecases,amethodmayonlysolveaweakerformofatask,anditisthereforeimportantforabrokeringagenttobeabletoreasonaboutthepetencematching,todecidewhetheritisOKtousetheweakermethodinthegivenscenario.Forinstance,inacurrencyconversionscenario,ataskspecificationmaydefinecurrencyconversionratesintermsoftheofficialFTquotes,butdifferentserviceprovidersmayadoptotherconversionrates.Byexplicitlydistinguishingbetweentasksandmethodsweprovideabasicframeworkforrepresentingthesedifferencesandforenablingmatchmakingagents[15]toreasonaboutthem. Theseparationbetweentasksandmethodsalsoprovidesabasicmodelfordealingwithontologymismatches.WhileinDAML-SsubscribingtoaServiceClassimpliesastrongmitment(i.e.,itmeanstodefinethenewserviceasaninstanceoftheclass),theUPMLframeworkassumesthatthemappingbetweenmethodsandtasksmaybemediatedbybridges.InpracticethismeansthatiftaskTisspecifiedinontologyAandamethodMisspecifiedinontologyB,whichcanbeusedtosolveT,itisstillpossibletouseMtosolveT,providedtheappropriatebridgeisdefined. Finally,anotheradvantageofourapproachisthatthetask-methoddistinctionalsoenablescapability-drivenserviceinvocation.WhilethisisalsopossibleinprincipleinapproachessuchasDAML-S,asdiscussedabove,ourapproachprovidesbothanexplicitseparationbetweenservicetypesandserviceprovidersandmoreflexibilityintheassociationbetweenmethodsandtasks. 3IRS–IIArchitecture TheoverallarchitectureoftheIRS–IIisshowninfigure2.TheponentsaretheIRSServer,theIRSPublisherandtheIRSClient,municatethroughaSOAP-basedprotocol[13]. 3.1IRSServer TheIRSserverholdsdescriptionsofsemanticwebservicesattwodifferentlevels.AknowledgeleveldescriptionisstoredusingtheUPMLframeworkoftasks,PSMsanddomainmodels.ThesearecurrentlyrepresentedinternallyinOCML[10],anOntolingua-derivedlanguagewhichprovidesboththeexpressivepowertoexpresstaskspecificationsandpetencies,aswellastheoperationalsupporttoreasonaboutthese.OnceruleandconstraintlanguagesaredevelopedforOWL[12], 310E.Mottaetal. wewillprovidetheappropriateimport/exportmechanisms.Inadditionwehavealsospecial-purposemappingmechanismstopetencespecificationstospecificwebservices.ThesecorrespondtothenotionofgroundinginDAML-
S. Fig.2.TheIRS–IIarchitecture 3.2TaskDescriptions Anexampletaskdescription,exchange_rate_provision,isshowninfigure3.Ascanbeseeninthefigurethetaskhastwoinputroles,asource_currencyandatarget_currency,andoneoutputrole,theexchange_rate.Thesupportingdefinitions,suchascurrencyandpositive_number,aredefinedinthetaskontologyassociatedwiththistask,orinontologiesincludedbyit. (def-classexchange_rate_provision(goal-specification-task)?
task((has-input-role:valuehas_source_currency:valuehas_target_currency)(has-output-role:valuehas_exchange_rate)(has_source_currency:typecurrency:cardinality1)(has_target_currency:typecurrency:cardinality1)(has_exchange_rate:typepositive_number) (has-goal-expression:value(kappa(?
psm?
sol)(=?
sol(the_official_exchange_rate(role-value?
psmhas_source_currency)(role-value?
psmhas_target_currency)))))))Fig.3.Definitionoftheexchange_rate_provisiontask Webservicemediationpositionaresupportedbytaskpreconditionsandgoalexpressions.Nopreconditionisspecifiedforthistask,althoughthespecificationsoftheinputrolesimplicitlystatethatone(andnomorethanone)sourceandtargetcurrencyhavetobespecified.Thegoalexpressionstatesthatthe IRS–II:AFrameworkandInfrastructureforSemanticWebServices311 outputforthetaskmustpliantwiththe“officialexchangerate”,asspecifiedintherelevantontology. (def-irs-soap-bindingsexchange_rate_provision_ontologyexchange_rate_provision((has-source-currency"xsd:symbol")(has-target-currency"xsd:symbol"))"xsd:float") ;;ontologyname;;taskname;;sourcecurrency;;targetcurrency;;output Fig.4.Thesoap-bindingsfortheexchange_rate_provisiontask TheintegrationofsemanticspecificationsandwebservicedescriptionsisachievedatthetasklevelbymeansofSOAPbindings.ASOAPbindingmapstheinputandoutputrolesontoSOAPtypes-thesoapbindingfortheexchange_rate_provisiontaskisshowninfigure4.Thebindingspecifiesthattheinputroles,source_currencyandtarget_currency,aremappedtotheSOAPtypexsd:symbolandtheoutputroleismappedtotheSOAPtypexsd:float.TherelationbetweenSOAPtypesandontologicalinputandoutputtypesisanalogoustothedistinctionbetweenknowledgeandsymbollevelinknowledgebasedsystems[11].Theontologyspecifieswhatknowledgeisrequiredandproduced;theSOAPtypesspecifythewaythisknowledgeiseffectivelyencodedinthemunicationmechanism.Hence,anywebservicewhichsolvesaparticulartaskplywithbothknowledgeandsymbollevelrequirements,oralternatively,bridgesneedtobedefinedtoensureinteroperability. 3.3ProblemSolvingMethods TheIRSserverholdsboththemethoddescriptions(PSMs)andaregistryofwebservices,whichimplementthem.AnexamplePSM,whichtacklestheexchange_rate_provisiontask,isshowninfigure5.Wecanseethatthetypeoftheinputroleshasbeenconstrainedfromcurrencytoeuropean_currency.Alsopreandpostconditionshavebeenintroduced. Thepreconditionstatesthatthebankmusthaveavailablestockofthetargetcurrency,whilstthepost-conditionstatesthattherateprovidedistheonesuppliedbytheEuropeanCentralBank(ECB).Hence,thisparticularservicemayormaynot‘solve’theexchangeprovisiontask,dependingonwhethertheexchangerateprovidedbyECBisthesameastheonerequiredbythetask,orwhetherthematchmakingagentishappytoconsiderthemas‘closeenough’.TheexplicitdistinctionbetweentasksandPSMsmakesitpossibletopreciselyspecify,bymeansofontologies,boththeproblemstobeaddressedandthedifferentwaystoaddressthemandprovidesabasistomatchmakingagentstoreasonaboutthemethod-to-totaskmappingandtomediationservicestotryand‘bridgethegap’betweenservicerequirementsandserviceproviders. Inasimilarfashiontotasks,webservicediscoveryissupportedbythepreandpostconditions.Forinstance,theconditionsformulatedintheMM_Bank_exchange_rate_providerPSMcanbeusedtoansweragentqueries 312E.Mottaetal. suchas“whichexchangerateservicesfocusonEuropeancurrencies”and“whichexchangerateservicesareabletochange250Kpoundsintoeuros?
”. (def-classMM_Bank_exchange_rate_provider(primitive-method)?
psm((has-input-role:valuehas-source-currency:valuehas-target-currency)(has-output-role:valuehas-exchange-rate)(has-source-currency:typeeuropean_currency:cardinality1)(has-target-currency:typeeuropean_currency:cardinality1)(has-exchange-rate:typepositive_number)(has-precondition:value(kappa(?
psm)(stock-available(role-value?
psmhas-target-currency))(has-postcondition:value(kappa(?
psm?
sol)(=?
sol(the-European-Central-Bank-exchange-rate(role-value?
psmhas-source-currency)(role-value?
psmhas-target-currency))))))) Fig.5.APSMwhichaddressestheexchange_rate_provisiontask 3.4IRSPublisher TheIRSPublisherplaystworolesintheIRS–IIframework.Firstly,itlinkswebservicestotheirsemanticdescriptionswithintheIRSserver.Notethatitispossibletohavemultipleservicesdescribedbythesamesemanticspecifications(i.e.,multipleimplementationofthesamefunctionality),aswellasmultiplesemanticspecificationsofthesameservice.Forinstance,thesameexchangerateconvertercanbedescribedusingtwodifferentontologiesforthefinancialsector.Secondly,itautomaticallygeneratesasetofwrapperswhichturnstandaloneLisporJavacodeintoawebservicedescribedbyaPSM.StandalonecodewhichispublishedontheIRSappearsasastandardjavawebservice.Thatis,awebserviceendpointisautomaticallygenerated. WebservicescanbepublishedusingeithertheIRSJavaAPIorthePublisherformbasedinterface.Figure6showsanIRS–IIuserpublishingawebservicewhichimplementstheMM_Bank_exchange_rate_provider.Asitcanbeseenfromthefigure,publishingastandardwebservicethroughtheIRSisveryeasy.Allthewebservicedeveloperhastodotois:
1.SpecifythelocationoftheIRSserverviaahostandportnumber.2.IndicatethePSMimplementedbytheservicebyprovidingitsnameandontology. ThemenuofavailablePSMsisgeneratedautomaticallyoncethelocationoftheIRSserverhasbeenspecified.3.Specifytheendpointforthewebservice.Ifthe‘service’inquestionisapieceofjavacode,specifiedas,thentheappropriatewrapperandanend-pointareautomaticallygeneratedbytheIRSpublisher.
IRS–II:AFrameworkandInfrastructureforSemanticWebServices313
Oncethe‘PublishWebService’buttonhasbeenpressed,aSOAPmessageencodingtheinformationintheformissenttotheIRSserverwhereanassociationbetweenthePSMandthewebserviceendpointisstoredintheregistryofimplementers.AJavaAPI,whichreplicatesthefunctionalityoftheform,alsoexists.
Fig.6.TheIRS–IIformbasedinterfaceforpublishingawebservice
AswementionedearliertheIRSPublisheralsoallowsstandaloneJavaandLispcodetobeturnedintoawebserviceandassociatedwithaPSMthroughasimpleAPI.ForLispamacroirs-method-registrationisused-anexamplefortheMM_Bank_exchange_rate_providerPSMisgiveninfigure7.Whentheforminfigure7isexecuted,asetofwrappersaregeneratedwhichmakethefunctionmmexchange-rateavailableasawebservice.ExecutingasecondIRSform(publish-all-services)sendsthedescriptionandlocationofallthenewlycreatedwebservicestotheIRSserver.TheIRSServerautomaticallygeneratesanendpoint,whichenablestheLispfunctiontobeessedasastandardwebservice.
(irs-method-registration
MM_Bank_exchange_rate_provider_ontology;;theontology
MM_Bank_exchange_rate_provider
;;thePSM
mm-exchange-rate)
;;theLispfunction
Fig.7.Registeringthelispfunctionmm-exchange-rateasanimplementationoftheMM_Bank_exchange_rate_providerPSM.
AsimilarAPIisprovidedforJava.Figure8belowshowshowaJavamethodimplementingtheMM_Bank_exchange_rate_providerPSMcouldbepublishedthroughtheIRSpublisher.
314E.Mottaetal.
IRSPublisherirsPublisher=newIRSPublisher("http://137.108.24.248:3000/soap");
//IRSserverURL
irsPublisher.PublishPSM("MM_Bank_exchange_rate_provider”,"MM_Bank_exchange_rate_provider_ontology","MM_Bank","exchangeRate”);
//PSMName//PSMOntology//Classname//methodname
Fig.8.TheexchangeRatemethodoftheJavaclassMM_bankpublishedasanimplementationoftheMM_Bank_exchange_rate_providerPSMthroughtheIRSPublisher.
3.5IRSClient
AkeyfeatureofIRS–IIisthatwebserviceinvocationiscapabilitydriven.TheIRSClientsupportsthisbyprovidinganinterfaceandasetofAPIswhicharetaskcentric.AnIRS–IIusersimplyasksforatasktobeachievedandtheIRS–IIbrokerlocatesanappropriatePSMandtheninvokesthecorrespondingwebservice-seesection4foranexample.Thesamefunctionalitycanalsobeinvokedprogrammatically,throughappropriateAPIsassociatedwiththecurrentclientplatforms,currentlyLispandJava.
4ThePatientShippingHealthcareScenario
ToillustratehowtheIRScanbeusedtodevelopapplicationsintermsofanumberofco-operating,distributedsemanticwebservices,wewilldescribeascenariotakenfromthehealth-caredomain.ThisscenariocoversaUKhealthcarepolicywhichwasintroducedin2002.Thepolicywastoreducewaitinglistsforthetreatmentofsomenon-urgentmedicalproblemsbygivingpatientswhowereexpectedtowaitmorethan6monthsforanoperationtheoptiontobetreatedinmainlandEurope.Figure9graphicallyillustrateshowwehaveimplementedthescenario,whichwedub“patientshipping”,withintheIRS–II.Tolimitthescopeoftheapplicationwefocusedonthemedicalconditionofarthritiswhichcansometimesrequiresurgery.
Ascanbeseeninfigure9fivemaintypesofwebservicesaresupported.Startingfromleftofthefigureandproceedingclockwisetheseare:•Adiagnosticandmenderserviceabletodiagnoseacondition,forexamplea
typeofarthritis,fromasetofsymptoms,andtomendatherapysuchasaparticularkindofsurgery.•AyellowpagesserviceabletoindicatewhichhospitalsaroundEuropeperformspecificmedicalservices.•Servicesassociatedwithindividualhospitalsabletoanswerqueriesabouttheavailabilityandcostofthespecificmedicaltreatmentstheyoffer.•Ambulanceservicesabletoprovidepricesforshippingpatientsfromonehospitaltoanotheracrossinternationalboundaries.•Anexchangerateserviceforconvertingpricesintolocalcurrencies.
IRS–II:AFrameworkandInfrastructureforSemanticWebServices315
Fig.9.AgraphicaloverviewofthepatientshippingscenarioTaskandPSMdescriptionswerecreatedfortheaboveserviceswithintheIRSserver,usingourknowledgemodellingtool,WebOnto[4].TheserviceswerethenimplementedinamixtureofJavaWebServices(exchangerate,ambulanceservicesandanumberofthehospitals)andLisp(alltheremaining),andpublishedusingtheIRSPublisher.Finally,apatientshippingwebservicewhichintegratestheaboveserviceswasimplementedandpublished.Thepatientshippingtaskhasfiveinputroles.Thefirstfourarethesymptomswhichthepatientdisplaysandthefifthisthelocationofthepatient.Figure10showsavisualizationofthedistributedapplicationduringtheexecution.ThevisualizationposedoftwocolumnsshowingtheIRSserverandelevenpublishedservices.Eachpublishedwebserviceisdisplayedinapanelcontaininga)thenameofthePSM,b)aniconicrepresentationofthestatusofthewebservice,andc)alogofthemessagesthewebservicesendsandreceives.Themeaningsoftheiconsare:
-thewebserviceiscurrentlyidle.-thewebserviceiscurrentlyprocessing.-thewebserviceissendingamessage.Wecanseeinfigure10thatanumberofserviceshavebeencalledwiththefollowingresults:•ThepatienthasbeendiagnosedwithsevereosteoarthritisbytheArthritisDiagnosis-Service.•TheArthritis-Therapy-ServicemendsthatthepatientistreatedbymeansofArthroplasty,asynonymforhip-replacement.
316E.Mottaetal.
Fig.10.Avisualizationofthepatientshippingwebserviceinmidexecution
•TheMedical_Service_Locatorservicehasfoundthreehospitalswhichofferhip-replacementasamedicalservice,specificallyAnother-Hippy-Hospital,The-Hippy-Hospital,andtheHip-Hip-Hospital.
•TheHip-Hip-Hospitalcantreatthepatientfirst(onthe20thofJune,2003).•TheAir_Ambulance_ServicecanmovethepatientfromMiltonKeynesto
Paris,thelocationoftheHip-Hip-Hospital,for3000Euros.
Wecanalsoseefromfigure10thatthreewebservicesarecurrentlyrunning:the
Shipping-Patient-Service;theGeneric_Currency_Converterandthe
Exchange-Rate-Provider.TheIRSserverhasjustsentamessagetothe
Exchange-Rate-ProviderrequestinganexchangeratebetweentheEurosource
currencyandthePoundtargetcurrency.Threemorestepswillurbeforethe
applicationterminates.First,theExchange-Rate-Providerwillsendanexchange
rate
to
the
Generic_Currency_Converter.
Second,
the
Generic_Currency_Converterwillconvertthe3000Eurosto1920pounds.
Finally,theShipping-Patient-Servicewillsendtheresultbacktotheclient
interface(showninfigure10).
Thisapplicationillustratessomeoftheadvantagesofsemanticwebservicesin
generalandourapproachinparticular.Servicediscoveryiscarriedoutusingsemantic
descriptions.Forinstance,onceaneedforhipreplacementhasbeenascertained,the
appropriatehospitalsareidentified,whichcanprovidehipreplacement,usinga
directoryofhospitalsandinterrogatingeachhospitalagentinturn.Thankstothe
availabilityofsemanticdescriptions,itisnotnecessarytoinvokehospitalweb
servicesdirectly.Instead,asemanticqueryforhospitalsprovidinghipreplacement
servicesissenttotheIRSandtheIRSbrokeristhenabletomatchthisqueryagainst
IRS–II:AFrameworkandInfrastructureforSemanticWebServices317
thesemanticdescriptionsofthevarioushospitalserviceproviders.Theotherimportantaspectistheuseofcapability-drivenserviceinvocation.Forinstance,onceahospitalhasbeenidentified,whichcantreatthepatientinParis,theapplicationclientsimplysendsan“achieve-task”messagetotheIRSserver,askingthelattertofindthecheapestproviderofambulanceservicesbetweenMiltonKeynesandParis.
5RelatedWork
TheframeworkusedbytheIRS–IIhasmuchmonwiththeWebServiceModellingFramework(WSMF)[5],asboththeIRS–IIandWSMFbuildonresearchinknowledgemodellingandinparticularontheUPMLframework.Asaresultbothapproachesemphasizetheimportanceofseparatinggoalandservicedescriptionstoensureflexibilityandscalability.ThemaindifferencebetweenIRS–IIandWSMFisthatwhilethelatterisexclusivelyaframework,theIRS–IIisalsoanimplementedinfrastructure,providingpublishingsupport,clientAPIs,brokeringandregistrymechanisms.
TheIRS–IIalsodiffersfromtheDAML-Sworkinanumberofways,asalreadydiscussedinsection2.Inparticular,DAML-Sdoesnotincludeflexibletasks-tomethodsmappingsandreliesinsteadonhierarchiesofservices,thuslimitingthepossibilitiesforflexible,n:mmediationbetweenproblemsandservices.Indeednoservice-independentnotionofproblemtypeispresentinDAML-
S.AnotherdifferenceisthatIRS–IIrepresentsdescriptionsinOCML,whileDAML-SusesDAML+OIL.Thisislikelytobeatemporarydifference,giventhatbothapproachesplantomovetoOWL-basedrepresentationsinthenearfuture. RegardingW3CWebServicesstandards,therearedifferencesintheapproachwetaketowardsapplicationdevelopmentandintherolesofponents.Forexample,unlikeUDDIregistries,whenaservicedescriptionispublishedtoIRS,thecodeforserviceinvocationisautomaticallygeneratedandlaterusedduringtaskachievement. 6Conclusions InthispaperwehavedescribedIRS–II,aframeworkandaninfrastructurewhichsupportsthepublication,discovery,positionanduseofsemanticwebservices.IRS–IIprovidesone-clickpublishingsupportfordifferentsoftwareplatforms,tofacilitatepublishingandsemanticannotationofwebservices.LikeWSMF,IRS–IIcapitalizesonknowledgemodellingresearchandisbasedonaflexibleframeworkseparatingservicefromproblemspecifications. FutureworkonIRS–IIwillimproveerrorhandling,whichatthemomentisverybasic.Wealsowanttofacilitateautomaticmediation,inordertoexploittheseparationoftasksandmethodsmorefully.Anotherimportantgoalistomoveawayfromabuilt-inmatchmakingfacilityandgeneralizethistoamatchmakinginfrastructure,essentiallyprovidinghooksfordifferentmatchmakingapproachestobeintegrated.Finally,weplantoOWL-ifytheinfrastructure,toensurepliancewithemergingsemanticwebstandards. 318E.Mottaetal. Acknowledgements.ThisworkhasbeenpartiallysupportedbytheAdvancedKnowledgeTechnologies(AKT)InterdisciplinaryResearchCollaboration(IRC),whichissponsoredbytheUKEngineeringandPhysicalSciencesResearchCouncilundergrantnumberGR/N15764/01.TheAKTprisestheUniversitiesofAberdeen,Edinburgh,Sheffield,SouthamptonandtheOpenUniversity. References
1.Benjamins,
V.R.,Plaza,
E.,Motta,
E.,Fensel,
D.,Studer,
R.,Wielinga,
B.,Schreiber,
G.,Zdrahal,
Z.,andDecker,
S.AnintelligentbrokeringserviceforponentreuseontheWorld-WideWeb.InGaines,B.andMusen,
M.(Editors),11thWorkshoponKnowledgeAcquisition,ModelingandManagement,Banff,Canada,1998.http://ksi.cpsc.ucalgary.ca/KAW/KAW98/benjamins3/
2.ClanceyW.J.(1985).HeuristicClassification.ArtificialIntelligence,27,pp.289–350.3.DAML-S0.7DraftRelease(2002).DAMLServicesCoalition..Availableonlineat /services/daml-s/0.7/.4.Domingue,
J.(1998)TadzebaoandWebOnto:Discussing,Browsing,andEditing OntologiesontheWeb.11thKnowledgeAcquisitionforKnowledge-BasedSystemsWorkshop,April18th–23rd.Banff,Canada.5.Fensel,
D.,Bussler,
C.(2002).TheWebServiceModelingFrameworkWSMF.AvailableatessProtocol(SOAP)(2000).W3CNote08.Availableonlineat/TR/SOAP/.14.McIlraith,
S.,Son,
T.C.,andZeng,
H.SemanticWebServices.IEEEIntelligentSystems,Mar/Apr.2001,pp.46–53.15.Sycara,
K.,Lu,
J.,Klusch,M.andWidoff,
S.MatchmakingamongHeterogeneousAgentsonthe.Proceedingsofthe1999AAAISpringSymposiumonIntelligentAgentsinCyberspace,StanfordUniversity,USA,22–24March1999.16.UDDISpecification.Availableonlineat17.WebServicesDescriptionLanguage(WSDL)(2001).W3CNote15.Availableonlineat/TR/wsdl
E.Motta,
J.B.Domingue,
L.S.Cabral}@open.ac.uk 2DipartimentodiScienzedell’Informazione,UniversityofBologna,Italygaspari@cs.unibo.it Abstract.InthispaperwedescribeIRS–II(ReasoningService)aframeworkandimplementedinfrastructure,whosemaingoalistosupportthepublication,location,positionandexecutionofheterogeneouswebservices,augmentedwithsemanticdescriptionsoftheirfunctionalities.IRS–IIhasthreemainclassesoffeatureswhichdistinguishitfromotherworkonsemanticwebservices.Firstly,itsupportsone-clickpublishingofstandalonesoftware:IRS–IIautomaticallycreatestheappropriatewrappers,givenpointerstothestandalonecode.Secondly,itexplicitlydistinguishesbetweentasks(whattodo)andmethods(howtoachievetasks)andasaresultsupportscapabilitydrivenserviceinvocation;flexiblemappingsbetweenservicesandproblemspecifications;anddynamic,knowledge-basedserviceselection.Finally,IRS–IIservicesarewebpatible–standardwebservicescanbetriviallypublishedthroughtheIRS–IIandanyIRS–IIserviceautomaticallyappearsasastandardwebservicetootherwebserviceinfrastructures.InthepaperweillustratethemainfunctionalitiesofIRS–IIthroughascenarioinvolvingadistributedapplicationinthehealthcaredomain. 1Introduction Webservicespromisetoturnthewebofstaticdocumentsintoavastlibraryofinteroperableputerprogramsandassuchhaveattractedconsiderableinterest,bothfromindustryandacademia.Forexample,IDC[8]predictsthattheWebServicesmarket,valuedat$416millionin2002,willbeworth$2.9billionby2006. Existingwebservicetechnologiesarebasedonamanualapproachtotheircreation,maintenanceandmanagement.Atthecentreoftheconceptualarchitectureisaregistrywhichstoresdescriptionsofpublishedwebservices.Clientsquerytheregistrytoobtainrelevantdetailsandtheninteractdirectlywiththedeployedservice.Thedescriptions,representedinXMLbaseddescriptionlanguages,suchasWSDL[17]andUDDI[16],mostlyfocusonthespecificationoftheinputandoutputdatatypesandtheessdetails.Thesespecificationsareobviouslynotpowerfulenoughtosupportautomaticdiscovery,mediationpositionofwebservices.Asoftwareagentcannotfindoutwhatawebserviceactuallydoes,byreasoningaboutaWSDLspecification.Analogouslythesameagentcannotlocatetheappropriate
D.Fenseletal.(Eds.):ISWC2003,LNCS2870,pp.306–318,2003.©Springer-VerlagBerlinHeidelberg2003 IRS–II:AFrameworkandInfrastructureforSemanticWebServices307 serviceinaUDDIregistry,giventhespecificationofatargetfunctionality.Asaresult,existingwebserviceinfrastructuresbyandlargesupportamanualapproachtowebservicemanagement:onlymanualdiscoveryissupportedandonly‘static’,manuallyconfiguredwebapplicationsarepossible. Theaboveissuesarebeingaddressedbyongoingworkintheareaofsemanticwebservices[3,5,14].Thebasicideahereisthatbyaugmentingwebserviceswithrichformaldescriptionsofpetencemanyaspectsoftheirmanagementwilleautomatic.Specifically,webservicelocation,positionandmediationcanedynamic,withsoftwareagentsabletoreasonaboutthefunctionalitiesprovidedbydifferentwebservices,abletolocatethebestonesforsolvingaparticularproblemandabletoposetherelevantwebservicestobuildapplicationsdynamically.Researchintheareaisrelativelynewandalthoughanumberofapproacheshavebeenproposed,suchasDAML-S[3]andWSMF[5],prehensivetoolinfrastructuresexist,whichsupportthespecificationanduseofsemanticwebservices. InthispaperwedescribeIRS–II(ReasoningService)aframeworkandimplementedinfrastructurewhichsupportsthepublication,location,positionandexecutionofheterogeneouswebservices,augmentedwithsemanticdescriptionsoftheirfunctionalities.IRS–IIhasthreemainclassesoffeatureswhichdistinguishitfromotherworkonsemanticwebservices. Firstly,itsupportsone-clickpublishingof‘standard’programmingcode.Inotherwords,itautomaticallytransformsprogrammingcode(currentlywesupportJavaandLispenvironments)intoawebservice,byautomaticallycreatingtheappropriatewrappers.Hence,itisveryeasytomakeexistingstandalonesoftwareavailableon,aswebservices. Secondly,theIRS–IIbuildsonknowledgemodelingresearchonponentsforknowledge-basedsystems[2,6,9,10],andasaresult,itsarchitectureexplicitlyseparatestaskspecifications(theproblemswhichneedtobesolved),fromthemethodspecifications(waystosolveproblems),fromthedomainmodels(wheretheseproblems,whichcanbegenericinnature,needtobesolved).Asaconsequence,IRS–IIisabletosupportcapability-drivenserviceinvocation(findmeaservicethatcansolveproblemX).Moreover,thecleandistinctionbetweentasksandmethodsenablesthespecificationofflexiblemappingsbetweenservicesandproblemspecifications,thusallowingan:mmappingbetweenproblemsandmethodsandadynamic,knowledge-basedserviceselection. Finally,IRS–IIservicesarewebpatible–standardwebservicescanbetriviallypublishedthroughtheIRS–IIandanyIRS–IIserviceautomaticallyappearsasastandardwebservicetootherwebserviceinfrastructures. Thepaperanizedasfollows:inthefollowingsectionweoutlineouroverallapproach.WethendescribetheIRS–IIframeworkindetailandillustrateitsponentsthroughascenarioinvolvingadistributedhealthcareapplication.Thefinalsectionofthepapercontainsourconclusions. 2TheIRS–IIApproach WorkontheIRS–IIbeganinthecontextoftheIBROWproject[1],whoseoverallgoalwastosupporton-the-flyapplicationdevelopmentthroughtheautomatic 308E.Mottaetal. configurationofreusableponents,availablefromdistributedlibrariesonthe.TheselibrariesarestructuredordingtotheUPMLframework[6],whichisshowninfigure1.TheUPMLframeworkdistinguishesbetweenthefollowingclassesponents:•Domainmodels.Thesedescribethedomainofanapplication(e.g.vehicles,a medicaldisease).•Taskmodels.Theseprovideagenericdescriptionofthetasktobesolved, specifyingtheinputandoutputtypes,thegoaltobeachievedandapplicablepreconditions.Taskscanbehigh-levelgenericdescriptionsplexclassesofapplications,suchasClassificationorScheduling,aswellasmore‘mundane’problemspecifications,suchasExchangeRateConversion.•ProblemSolvingMethods(PSMs).Theseprovideabstract,implementationindependentdescriptionsofreasoningprocesseswhichcanbeappliedtosolvetasksinspecificdomains.Asinthecaseoftaskmodels,thesecanbehigh-level,genericPSMssuchasHeuristicClassification[2]andPropose&Revise[9],ortheycanbespecializedmethodsapplicabletofine-grainedtasks,suchasExchangeRateConversion.•Bridges.Thesespecifymappingsbetweenthedifferentponentswithinanapplication.Forexample,therefinementprocessinheuristicclassificationmaybemappedontoataxonomichierarchyofattributeswithinsomedomain,inordertoconstructaspecificapplication.Eachclassponentisspecifiedbymeansofanappropriateontology[7].Themainadvantageofthisframeworkfromanepistemologicalpointofviewisthatitclearlyseparatestheponentsofknowledge-basedapplications,thusprovidingatheoreticalbasisforanalyzingknowledge-basedreasonersandanengineeringbasisforstructuringlibrariesofponents,performingknowledgeacquisition,andcarryingoutapplicationdevelopmentbyreuse[10]. Fig.1.TheUPMLframework TheapplicationoftheUPMLframeworktosemanticwebservicesalsoprovidesanumberofadvantagesandinourviewourparesfavorablywithapproachessuchasDAML-S,whereservicesarearrangedinhierarchiesandnoexplicitnotionoftaskisprovided–tasksaredefinedasservice-seekingagents.In IRS–II:AFrameworkandInfrastructureforSemanticWebServices309 DAML-Stasksarealwaysapplicationspecific,noprovisionfortaskregistriesisenvisaged.Incontrastinourapproach,tasksprovidethebasicmechanismforaggregatingservicesanditispossibletospecifyservicetypes(i.e.,tasks),independentlyofspecificserviceproviders.InprinciplethisisalsopossibleinDAML-
S.Hereataskwouldbedefinedasaserviceclass,sayS,anditsprofilewillgivethetaskdefinition.However,thissolutionimpliesthatallinstancesofSwillinheritthetaskprofile.Thisapproachisnotveryflexible,giventhatitmakesitimpossibletodistinguish(andtoreasonabout)thedifferencesbetweentheprofileofatask(serviceclass)andtheprofileofamethod(specificserviceprovider)–attributesareinheriteddownis-ahierarchies.Inparticular,insomecases,amethodmayonlysolveaweakerformofatask,anditisthereforeimportantforabrokeringagenttobeabletoreasonaboutthepetencematching,todecidewhetheritisOKtousetheweakermethodinthegivenscenario.Forinstance,inacurrencyconversionscenario,ataskspecificationmaydefinecurrencyconversionratesintermsoftheofficialFTquotes,butdifferentserviceprovidersmayadoptotherconversionrates.Byexplicitlydistinguishingbetweentasksandmethodsweprovideabasicframeworkforrepresentingthesedifferencesandforenablingmatchmakingagents[15]toreasonaboutthem. Theseparationbetweentasksandmethodsalsoprovidesabasicmodelfordealingwithontologymismatches.WhileinDAML-SsubscribingtoaServiceClassimpliesastrongmitment(i.e.,itmeanstodefinethenewserviceasaninstanceoftheclass),theUPMLframeworkassumesthatthemappingbetweenmethodsandtasksmaybemediatedbybridges.InpracticethismeansthatiftaskTisspecifiedinontologyAandamethodMisspecifiedinontologyB,whichcanbeusedtosolveT,itisstillpossibletouseMtosolveT,providedtheappropriatebridgeisdefined. Finally,anotheradvantageofourapproachisthatthetask-methoddistinctionalsoenablescapability-drivenserviceinvocation.WhilethisisalsopossibleinprincipleinapproachessuchasDAML-S,asdiscussedabove,ourapproachprovidesbothanexplicitseparationbetweenservicetypesandserviceprovidersandmoreflexibilityintheassociationbetweenmethodsandtasks. 3IRS–IIArchitecture TheoverallarchitectureoftheIRS–IIisshowninfigure2.TheponentsaretheIRSServer,theIRSPublisherandtheIRSClient,municatethroughaSOAP-basedprotocol[13]. 3.1IRSServer TheIRSserverholdsdescriptionsofsemanticwebservicesattwodifferentlevels.AknowledgeleveldescriptionisstoredusingtheUPMLframeworkoftasks,PSMsanddomainmodels.ThesearecurrentlyrepresentedinternallyinOCML[10],anOntolingua-derivedlanguagewhichprovidesboththeexpressivepowertoexpresstaskspecificationsandpetencies,aswellastheoperationalsupporttoreasonaboutthese.OnceruleandconstraintlanguagesaredevelopedforOWL[12], 310E.Mottaetal. wewillprovidetheappropriateimport/exportmechanisms.Inadditionwehavealsospecial-purposemappingmechanismstopetencespecificationstospecificwebservices.ThesecorrespondtothenotionofgroundinginDAML-
S. Fig.2.TheIRS–IIarchitecture 3.2TaskDescriptions Anexampletaskdescription,exchange_rate_provision,isshowninfigure3.Ascanbeseeninthefigurethetaskhastwoinputroles,asource_currencyandatarget_currency,andoneoutputrole,theexchange_rate.Thesupportingdefinitions,suchascurrencyandpositive_number,aredefinedinthetaskontologyassociatedwiththistask,orinontologiesincludedbyit. (def-classexchange_rate_provision(goal-specification-task)?
task((has-input-role:valuehas_source_currency:valuehas_target_currency)(has-output-role:valuehas_exchange_rate)(has_source_currency:typecurrency:cardinality1)(has_target_currency:typecurrency:cardinality1)(has_exchange_rate:typepositive_number) (has-goal-expression:value(kappa(?
psm?
sol)(=?
sol(the_official_exchange_rate(role-value?
psmhas_source_currency)(role-value?
psmhas_target_currency)))))))Fig.3.Definitionoftheexchange_rate_provisiontask Webservicemediationpositionaresupportedbytaskpreconditionsandgoalexpressions.Nopreconditionisspecifiedforthistask,althoughthespecificationsoftheinputrolesimplicitlystatethatone(andnomorethanone)sourceandtargetcurrencyhavetobespecified.Thegoalexpressionstatesthatthe IRS–II:AFrameworkandInfrastructureforSemanticWebServices311 outputforthetaskmustpliantwiththe“officialexchangerate”,asspecifiedintherelevantontology. (def-irs-soap-bindingsexchange_rate_provision_ontologyexchange_rate_provision((has-source-currency"xsd:symbol")(has-target-currency"xsd:symbol"))"xsd:float") ;;ontologyname;;taskname;;sourcecurrency;;targetcurrency;;output Fig.4.Thesoap-bindingsfortheexchange_rate_provisiontask TheintegrationofsemanticspecificationsandwebservicedescriptionsisachievedatthetasklevelbymeansofSOAPbindings.ASOAPbindingmapstheinputandoutputrolesontoSOAPtypes-thesoapbindingfortheexchange_rate_provisiontaskisshowninfigure4.Thebindingspecifiesthattheinputroles,source_currencyandtarget_currency,aremappedtotheSOAPtypexsd:symbolandtheoutputroleismappedtotheSOAPtypexsd:float.TherelationbetweenSOAPtypesandontologicalinputandoutputtypesisanalogoustothedistinctionbetweenknowledgeandsymbollevelinknowledgebasedsystems[11].Theontologyspecifieswhatknowledgeisrequiredandproduced;theSOAPtypesspecifythewaythisknowledgeiseffectivelyencodedinthemunicationmechanism.Hence,anywebservicewhichsolvesaparticulartaskplywithbothknowledgeandsymbollevelrequirements,oralternatively,bridgesneedtobedefinedtoensureinteroperability. 3.3ProblemSolvingMethods TheIRSserverholdsboththemethoddescriptions(PSMs)andaregistryofwebservices,whichimplementthem.AnexamplePSM,whichtacklestheexchange_rate_provisiontask,isshowninfigure5.Wecanseethatthetypeoftheinputroleshasbeenconstrainedfromcurrencytoeuropean_currency.Alsopreandpostconditionshavebeenintroduced. Thepreconditionstatesthatthebankmusthaveavailablestockofthetargetcurrency,whilstthepost-conditionstatesthattherateprovidedistheonesuppliedbytheEuropeanCentralBank(ECB).Hence,thisparticularservicemayormaynot‘solve’theexchangeprovisiontask,dependingonwhethertheexchangerateprovidedbyECBisthesameastheonerequiredbythetask,orwhetherthematchmakingagentishappytoconsiderthemas‘closeenough’.TheexplicitdistinctionbetweentasksandPSMsmakesitpossibletopreciselyspecify,bymeansofontologies,boththeproblemstobeaddressedandthedifferentwaystoaddressthemandprovidesabasistomatchmakingagentstoreasonaboutthemethod-to-totaskmappingandtomediationservicestotryand‘bridgethegap’betweenservicerequirementsandserviceproviders. Inasimilarfashiontotasks,webservicediscoveryissupportedbythepreandpostconditions.Forinstance,theconditionsformulatedintheMM_Bank_exchange_rate_providerPSMcanbeusedtoansweragentqueries 312E.Mottaetal. suchas“whichexchangerateservicesfocusonEuropeancurrencies”and“whichexchangerateservicesareabletochange250Kpoundsintoeuros?
”. (def-classMM_Bank_exchange_rate_provider(primitive-method)?
psm((has-input-role:valuehas-source-currency:valuehas-target-currency)(has-output-role:valuehas-exchange-rate)(has-source-currency:typeeuropean_currency:cardinality1)(has-target-currency:typeeuropean_currency:cardinality1)(has-exchange-rate:typepositive_number)(has-precondition:value(kappa(?
psm)(stock-available(role-value?
psmhas-target-currency))(has-postcondition:value(kappa(?
psm?
sol)(=?
sol(the-European-Central-Bank-exchange-rate(role-value?
psmhas-source-currency)(role-value?
psmhas-target-currency))))))) Fig.5.APSMwhichaddressestheexchange_rate_provisiontask 3.4IRSPublisher TheIRSPublisherplaystworolesintheIRS–IIframework.Firstly,itlinkswebservicestotheirsemanticdescriptionswithintheIRSserver.Notethatitispossibletohavemultipleservicesdescribedbythesamesemanticspecifications(i.e.,multipleimplementationofthesamefunctionality),aswellasmultiplesemanticspecificationsofthesameservice.Forinstance,thesameexchangerateconvertercanbedescribedusingtwodifferentontologiesforthefinancialsector.Secondly,itautomaticallygeneratesasetofwrapperswhichturnstandaloneLisporJavacodeintoawebservicedescribedbyaPSM.StandalonecodewhichispublishedontheIRSappearsasastandardjavawebservice.Thatis,awebserviceendpointisautomaticallygenerated. WebservicescanbepublishedusingeithertheIRSJavaAPIorthePublisherformbasedinterface.Figure6showsanIRS–IIuserpublishingawebservicewhichimplementstheMM_Bank_exchange_rate_provider.Asitcanbeseenfromthefigure,publishingastandardwebservicethroughtheIRSisveryeasy.Allthewebservicedeveloperhastodotois:
1.SpecifythelocationoftheIRSserverviaahostandportnumber.2.IndicatethePSMimplementedbytheservicebyprovidingitsnameandontology. ThemenuofavailablePSMsisgeneratedautomaticallyoncethelocationoftheIRSserverhasbeenspecified.3.Specifytheendpointforthewebservice.Ifthe‘service’inquestionisapieceofjavacode,specifiedas
S.AnotherdifferenceisthatIRS–IIrepresentsdescriptionsinOCML,whileDAML-SusesDAML+OIL.Thisislikelytobeatemporarydifference,giventhatbothapproachesplantomovetoOWL-basedrepresentationsinthenearfuture. RegardingW3CWebServicesstandards,therearedifferencesintheapproachwetaketowardsapplicationdevelopmentandintherolesofponents.Forexample,unlikeUDDIregistries,whenaservicedescriptionispublishedtoIRS,thecodeforserviceinvocationisautomaticallygeneratedandlaterusedduringtaskachievement. 6Conclusions InthispaperwehavedescribedIRS–II,aframeworkandaninfrastructurewhichsupportsthepublication,discovery,positionanduseofsemanticwebservices.IRS–IIprovidesone-clickpublishingsupportfordifferentsoftwareplatforms,tofacilitatepublishingandsemanticannotationofwebservices.LikeWSMF,IRS–IIcapitalizesonknowledgemodellingresearchandisbasedonaflexibleframeworkseparatingservicefromproblemspecifications. FutureworkonIRS–IIwillimproveerrorhandling,whichatthemomentisverybasic.Wealsowanttofacilitateautomaticmediation,inordertoexploittheseparationoftasksandmethodsmorefully.Anotherimportantgoalistomoveawayfromabuilt-inmatchmakingfacilityandgeneralizethistoamatchmakinginfrastructure,essentiallyprovidinghooksfordifferentmatchmakingapproachestobeintegrated.Finally,weplantoOWL-ifytheinfrastructure,toensurepliancewithemergingsemanticwebstandards. 318E.Mottaetal. Acknowledgements.ThisworkhasbeenpartiallysupportedbytheAdvancedKnowledgeTechnologies(AKT)InterdisciplinaryResearchCollaboration(IRC),whichissponsoredbytheUKEngineeringandPhysicalSciencesResearchCouncilundergrantnumberGR/N15764/01.TheAKTprisestheUniversitiesofAberdeen,Edinburgh,Sheffield,SouthamptonandtheOpenUniversity. References
1.Benjamins,
V.R.,Plaza,
E.,Motta,
E.,Fensel,
D.,Studer,
R.,Wielinga,
B.,Schreiber,
G.,Zdrahal,
Z.,andDecker,
S.AnintelligentbrokeringserviceforponentreuseontheWorld-WideWeb.InGaines,B.andMusen,
M.(Editors),11thWorkshoponKnowledgeAcquisition,ModelingandManagement,Banff,Canada,1998.http://ksi.cpsc.ucalgary.ca/KAW/KAW98/benjamins3/
2.ClanceyW.J.(1985).HeuristicClassification.ArtificialIntelligence,27,pp.289–350.3.DAML-S0.7DraftRelease(2002).DAMLServicesCoalition..Availableonlineat /services/daml-s/0.7/.4.Domingue,
J.(1998)TadzebaoandWebOnto:Discussing,Browsing,andEditing OntologiesontheWeb.11thKnowledgeAcquisitionforKnowledge-BasedSystemsWorkshop,April18th–23rd.Banff,Canada.5.Fensel,
D.,Bussler,
C.(2002).TheWebServiceModelingFrameworkWSMF.AvailableatessProtocol(SOAP)(2000).W3CNote08.Availableonlineat/TR/SOAP/.14.McIlraith,
S.,Son,
T.C.,andZeng,
H.SemanticWebServices.IEEEIntelligentSystems,Mar/Apr.2001,pp.46–53.15.Sycara,
K.,Lu,
J.,Klusch,M.andWidoff,
S.MatchmakingamongHeterogeneousAgentsonthe.Proceedingsofthe1999AAAISpringSymposiumonIntelligentAgentsinCyberspace,StanfordUniversity,USA,22–24March1999.16.UDDISpecification.Availableonlineat17.WebServicesDescriptionLanguage(WSDL)(2001).W3CNote15.Availableonlineat/TR/wsdl
声明：
该资讯来自于互联网网友发布，如有侵犯您的权益请联系我们。
相关文章
